Pith. sign in

REVIEW 5 major objections 6 minor 37 references

Template-Based Financial Report Generation in Agentic and Decomposed Information Retrieval

T0 review · 5 major / 6 minor · reviewed 2026-08-16 · deepseek-v4-flash

Pith's one-line read This paper claims that DecomposedIR, which breaks a report template into a chain of sub-questions answered with retrieved evidence, produces broader and more detailed templated reports than the multi-agent AgenticIR, with statistical…

desk verdict DecomposedIR likely beats AgenticIR for template-based report generation, but the headline significance claim rests on a length-confounded, self-evaluated metric and a table that contradicts it. read the letter →

arxiv 2504.14233 v1 pith:4VTRS6JE submitted 2025-04-19 cs.IR

classification cs.IR
keywords template-basedreportgenerationfinancialearningscallsummarizationagenticinformationretrievaldecomposedpromptingpromptchainingreference-freeevaluationlargelanguagemodels
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper asks which of two LLM workflows writes better reports when the output must follow a fixed multi-section template: an agentic system in which several agents collaborate with the whole template in view, or a decomposed system in which the template is broken into sub-questions and answered step by step with retrieved evidence. The authors build both pipelines for quarterly earnings releases from five semiconductor companies and evaluate the generated reports on four financial characteristics: financial takeaways, financial context, reasoning correctness, and management expectation. They report that the decomposed approach, DecomposedIR, statistically significantly outperforms the agentic approach on all four characteristics with p < 0.05, and that the same pattern holds on the SumIPCC climate-report dataset with expert-written references, where DecomposedIR beats AgenticIR by 33% in ROUGE-1 and 6.3% in BERTScore. The conclusion is that explicit template-to-query decomposition helps an LLM bridge the gap between unstructured source text and a structured template more reliably than autonomous agent collaboration. This matters for practitioners deciding whether to invest in multi-agent orchestration or use simpler prompt chaining for structured reporting.

What carries the argument

The mechanism that carries the argument is template-to-subquery decomposition via prompt chaining. DecomposedIR takes the analyst-defined report template, has an LLM convert each section into a specific sub-question, retrieves the top three transcript chunks plus structured financial statements for each sub-question, answers each one, and then aggregates the answers into a final report aligned with the template. AgenticIR, by contrast, presents the entire template to a team of agents equipped with task-decomposition and retrieval-generation functions, letting the agents choose how to split the work. The load-bearing difference is that decomposed sub-queries force the model to address every template section explicitly, whereas the agentic approach depends on the agents' autonomous decomposition staying aligned with the template.

What would settle it

Have financial analysts rate a matched sample of AgenticIR and DecomposedIR reports for the same earnings calls, or recompute DecompEval on length-matched subsamples of the reports; if human preference or length-matched scores do not reproduce the reported 27% advantage, the central claim fails.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that DecomposedIR outperforms AgenticIR for template-based financial report generation across all four key characteristics, on average by 27% with p < 0.05, and that the advantage transfers to a different domain: on SumIPCC, DecomposedIR scores 0.32 versus 0.24 in ROUGE-1 and 0.67 versus 0.63 in BERTScore. The authors interpret this as evidence that the agentic framework's flexible, autonomous division of labour is less able to align generated content with a structured template than a prompt-chaining workflow that reframes each template section as a retrieval-augmented question. They also find that adding a self-reflection step improves both approaches on the evaluation metrics but produces longer and more complex reports, and they note that agentic approaches may still be attractive for concise, orchestrated outputs. The paper frames the result as a reflection on when to use agentic frameworks in real-world applications.

Load-bearing premise

The load-bearing premise is that DecompEval and G-Eval measure real report quality rather than rewarding length, since DecomposedIR writes about 78 sentences per report versus 25 for AgenticIR and DecompEval adds up sentence-level hits, a confound the paper does not resolve.

Editorial extensions

If this is right

  • DecomposedIR is the stronger default for templated report generation when coverage of every template section matters, since it beat AgenticIR in two domains with different evaluation setups.
  • Self-reflection improves both frameworks on the four financial characteristics, so the two techniques can be combined rather than treated as exclusive alternatives.
  • The four-characteristic reference-free evaluation recipe (DecompEval and G-Eval) gives practitioners a no-gold-label way to compare report generators in the financial domain.
  • AgenticIR's weakness is attributed to missing explicit structural guidance, so adding template-structure constraints to agent workflows is a plausible way to close the gap.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: because DecomposedIR's reports are roughly three times longer (77.7 vs 25.0 sentences per report) and DecompEval aggregates sentence-level adherence, the 27% advantage could shrink substantially under length-controlled evaluation or independent human ratings.
  • Editorial inference: the same template-to-subquery recipe should transfer to other structured reporting tasks, such as regulatory filings, clinical notes, or equipment maintenance logs, where unstructured sources must fill a fixed schema; the paper does not test these domains.
  • Editorial inference: the results suggest reserving multi-agent architectures for tasks that genuinely need tool use, negotiation, or backtracking, and using linear prompt chaining for fixed-schema extraction; that is a cost-conscious reading the authors touch on but do not fully develop.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 6 minor

Summary. The paper compares two LLM-based workflows for generating templated financial reports from earnings releases: an AutoGen-based multi-agent framework (AgenticIR) and a decomposed prompt-chaining framework (DecomposedIR) in which each template section becomes a subquery answered with retrieved transcript chunks and financial statements. Experiments on 74 earnings documents from five semiconductor companies use reference-free DecompEval and G-Eval over four characteristics plus readability metrics; a second scenario uses the SumIPCC climate dataset with ROUGE/BERTScore against expert-written summaries. The authors report that DecomposedIR statistically significantly outperforms AgenticIR in both scenarios and that self-reflection improves metric scores while increasing text complexity.

Significance. The comparison addresses a practical question in applied LLM system design, and the paper is transparent about data and code availability. It also includes a cross-domain, reference-based check that goes in the same direction as the main result, which is a genuine strength. However, the headline statistical claim currently rests on unvalidated, length-sensitive LLM evaluation in the financial domain and on a very small reference-based dataset with no significance testing; these issues must be resolved before the comparative conclusion can be accepted.

major comments (5)
  1. [4.1, Tables 1 and 2] DecompEval aggregates sentence-level judgments, and DecomposedIR produces roughly three times more sentences (77.72 vs 25.01; 96.07 vs 31.82 with self-reflection). A longer report can therefore accumulate a higher DecompEval score even if its sentence-level density of relevant content is identical or worse. G-Eval's holistic scores are also sensitive to verbosity and detail volume. The paper does not report precision-style metrics such as the fraction of sentences judged on-topic, length-controlled comparisons, or human judgments, so the reported 27% average gain cannot be attributed to template coverage rather than output volume.
  2. [4.1, statistical test] The claim of statistical significance is not reproducible: Pearson's chi-square test is invoked with p<0.05, but no test statistic, degrees of freedom, p-value per characteristic, or unit of analysis is given. The reported numbers are average scores rather than raw counts, so it is unclear what contingency table was tested. Please specify the test construction and provide per-characteristic statistics; otherwise this central claim is unsupported.
  3. [4.2, Table 3] The cross-domain evidence comprises only seven reorganized reports, and no significance test, confidence interval, or variance measure is reported. With n=7 and differences of 0.08 in ROUGE-1 and 0.04 in BERTScore, the statement that DecomposedIR statistically significantly outperforms AgenticIR in both scenarios is not established. The reorganization of 140 summaries into seven multi-section reports also creates dependencies among evaluation units; please justify this aggregation and report report-level variance.
  4. [4.1, evaluation metrics] The financial-domain evaluation uses GPT-4o-mini both as the report generator and as the DecompEval/G-Eval judge, and the four characteristics are not validated against expert human annotation. This self-evaluation loop is particularly problematic because the judge may systematically favor the longer, more verbose outputs of DecomposedIR. At minimum, calibrate the LLM metrics on a small set of expert-rated reports or report inter-judge agreement with human financial analysts.
  5. [4.1, Table 1] The text states that DecomposedIR methods consistently outperform AgenticIR methods in all metrics, but Table 1 shows Financial Takeaways DE of 0.23 for DecomposedIR versus 0.26 for AgenticIR, and 0.27 for AgenticIR with self-reflection. This internal inconsistency either indicates a reporting error or shows that the per-characteristic advantage is not universal; it should be corrected and the claims qualified.
minor comments (6)
  1. [4.2] The word 'dataet' should be corrected to 'dataset'.
  2. [References [32] and [33]] AutoGen is cited twice with inconsistent reference details; please consolidate into a single consistent citation.
  3. [4.1] The phrase 'apple-to-apple comparison' should be 'apples-to-apples comparison'.
  4. [4.1, Table 2] Readability differences are reported without statistical tests; please state whether the observed differences are meaningful beyond sampling variation.
  5. [Abstract and Conclusion] The phrase 'statistically significantly outperforms' is used for both scenarios, but no significance test is described for the SumIPCC scenario; align the terminology with the analyses actually reported.
  6. [4, experimental setup] The retrieval depth n=3 is fixed without sensitivity analysis; since both methods depend on the retrieved chunks, a brief robustness check would strengthen the comparison.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the comparison is empirical, uses external metrics, and is independently anchored by the reference-based SumIPCC evaluation.

full rationale

The paper's central claim is an empirical comparison of two pipelines, not a derivation whose output is equivalent to its inputs. AgenticIR and DecomposedIR are run on the same data with the same LLM (GPT-4o-mini), the same embedding model per dataset, and the same retrieval depth n=3; no parameter is fitted to the reported outcome, and the headline result is not fed back into either method. The financial-domain evaluation uses the external reference-free metrics DecompEval and G-Eval, while the SumIPCC scenario uses reference-based ROUGE and BERTScore against human-written summaries, providing an independent check. There are no self-citations in the reference list, and the four evaluated characteristics are drawn from prior external work, not from the authors' own theorems. Two concerns are legitimate but do not amount to circularity under this pass's rules. First, DecompEval aggregates sentence-level adherence judgments and DecomposedIR produces roughly 77.7 sentences per report versus 25.0 for AgenticIR (Table 2), so output length could confound the aggregate DE scores; however, the paper does not define 'better coverage' as 'more sentences' by construction, and the reference-based SumIPCC result points in the same direction, so this is a measurement-validity threat rather than an equation-level reduction. Second, using GPT-4o-mini for both generation and evaluation creates a self-evaluation loop, but the evaluator is an external metric, not a quantity derived from the generation procedure, and no fitted quantity is renamed as a prediction. The paper itself acknowledges the absence of human references in Section 4.1: 'Due to the predominant cost of collecting expert-written template-based financial reports, it is not feasible to conduct an apple-to-apple comparison between the generated report and a gold reference.' A separate reporting inconsistency exists: Table 1 shows AgenticIR scoring higher than DecomposedIR on Financial Takeaways DE (0.26 versus 0.23), contradicting the text's claim that DecomposedIR 'consistently outperform[s] AgenticIR methods in all metrics'; this is an accuracy or reporting issue, not circular reasoning. Overall, the central comparison does not reduce to its inputs by construction or by self-citation.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The paper introduces no new formal entities or fitted physical parameters. The central comparison rests on domain assumptions about template validity, LLM-based evaluation, model choice, and dataset representativeness.

free parameters (2)
  • top-n retrieved chunks (n) = 3
    Uniformly set to 3 for both methods in Section 4. Chosen by hand, not fitted; affects the amount of context available for report generation and could influence both methods differently.
  • transcript chunk size and overlap = 1000 characters, 200-character overlap
    Preprocessing choice in Section 3.1; chosen by hand and could change retrieval quality for both methods.
assumptions (5)
  • domain assumption The financial report template obtained from an analyst is a valid representation of industry report structure.
    The template defines the sections and the subqueries in DecomposedIR (Section 3.1). If the template is atypical, the comparison may not generalize.
  • domain assumption LLM-based reference-free metrics (DecompEval and G-Eval) are valid proxies for financial report quality without human references.
    Section 4.1 adopts these metrics because expert-written reports are unavailable; their validity in the financial domain is not established.
  • domain assumption GPT-4o-mini is a representative LLM for both generation and evaluation.
    All experiments use GPT-4o-mini (Section 4); the comparison might differ with other LLMs, and using the same model for generation and evaluation introduces a self-evaluation loop.
  • domain assumption The five selected semiconductor companies and 74 earnings releases are representative of financial reporting tasks.
    Section 3.1 describes the dataset as a pilot exploration; broader financial sectors are not covered.
  • domain assumption Predefined embedding models provide adequate semantic retrieval for both methods.
    fin-mpnet-base and stella_en_1.5B_v5 are used for retrieval (Section 4); retrieval quality bounds both methods.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Template-Based Financial Report Generation in Agentic and Decomposed Information Retrieval." pith.science (2026). https://pith.science/paper/4VTRS6JE

@misc{pith2026250414233,
  author       = {Pith},
  title        = {Pith review of: Template-Based Financial Report Generation in Agentic and Decomposed Information Retrieval},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4VTRS6JE}},
  note         = {Machine review of arXiv:2504.14233}
}
read the original abstract

Tailoring structured financial reports from companies' earnings releases is crucial for understanding financial performance and has been widely adopted in real-world analytics. However, existing summarization methods often generate broad, high-level summaries, which may lack the precision and detail required for financial reports that typically focus on specific, structured sections. While Large Language Models (LLMs) hold promise, generating reports adhering to predefined multi-section templates remains challenging. This paper investigates two LLM-based approaches popular in industry for generating templated financial reports: an agentic information retrieval (IR) framework and a decomposed IR approach, namely AgenticIR and DecomposedIR. The AgenticIR utilizes collaborative agents prompted with the full template. In contrast, the DecomposedIR approach applies a prompt chaining workflow to break down the template and reframe each section as a query answered by the LLM using the earnings release. To quantitatively assess the generated reports, we evaluated both methods in two scenarios: one using a financial dataset without direct human references, and another with a weather-domain dataset featuring expert-written reports. Experimental results show that while AgenticIR may excel in orchestrating tasks and generating concise reports through agent collaboration, DecomposedIR statistically significantly outperforms AgenticIR approach in providing broader and more detailed coverage in both scenarios, offering reflection on the utilization of the agentic framework in real-world applications.

Figures

Figures reproduced from arXiv: 2504.14233 by the authors.

Figure 1
Figure 1. An example of the financial report template. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. An illustration of the AgenticIR framework. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 24 canonical work pages

  1. [1]

    Alex Yaw Adom, Israel Kofi Nyarko, and Gladys Narki Kumi Som. 2016. Com- petitor analysis in strategic management: Is it a worthwhile managerial practice in contemporary times. Journal of Resources Development and Management 24, 1 (2016), 116–127

  2. [2]

    Meri Coleman and Ta Lin Liau. 1975. A computer readability formula designed for machine scoring. Journal of Applied Psychology 60, 2 (1975), 283

  3. [3]

    Saurous, Jascha Sohl-Dickstein, Kevin Murphy, and Charles Sutton

    David Dohan, Winnie Xu, Aitor Lewkowycz, Jacob Austin, David Bieber, Raphael Gontijo Lopes, Yuhuai Wu, Henryk Michalewski, Rif A. Saurous, Jascha Sohl-Dickstein, Kevin Murphy, and Charles Sutton. 2022. Language Model Cas- cades. CoRR abs/2207.10342 (2022)

  4. [4]

    Kelvin Du, Frank Xing, Rui Mao, and Erik Cambria. 2024. Financial Sentiment Analysis: Techniques and Applications. ACM Comput. Surv. 56, 9 (2024), 220:1– 220:42

  5. [5]

    Gihan Gamage, Nishan Mills, Daswin De Silva, Milos Manic, Harsha Moraliyage, Andrew Jennings, and Damminda Alahakoon. 2024. Multi-Agent RAG Chatbot Architecture for Decision Support in Net-Zero Emission Energy Systems. In ICIT. IEEE, 1–6

  6. [6]

    Iacopo Ghinassi, Leonardo Catalano, and Tommaso Colella. 2024. Efficient Aspect- Based Summarization of Climate Change Reports with Small Language Models. CoRR abs/2411.14272 (2024)

  7. [7]

    Tomas Goldsack, Yang Wang, Chenghua Lin, and Chung-Chi Chen. 2025. From Facts to Insights: A Study on the Generation and Evaluation of Analytical Re- ports for Deciphering Earnings Calls. In COLING. Association for Computational Linguistics, 10576–10593

  8. [8]

    Pengcheng He, Jianfeng Gao, and Weizhu Chen. 2023. DeBERTaV3: Improv- ing DeBERTa using ELECTRA-Style Pre-Training with Gradient-Disentangled Embedding Sharing. In ICLR. OpenReview.net

Show all 37 references
  1. [9]

    Yi He, Wenxin Tai, Fan Zhou, and Yi Yang. 2023. Exploring Hypergraph of Earnings Call for Risk Prediction (Student Abstract). InAAAI. AAAI Press, 16226– 16227

  2. [10]

    Pranab Islam, Anand Kannappan, Douwe Kiela, Rebecca Qian, Nino Scherrer, and Bertie Vidgen. 2023. FinanceBench: A New Benchmark for Financial Question Answering. CoRR abs/2311.11944 (2023)

  3. [11]

    Pei Ke, Fei Huang, Fei Mi, Yasheng Wang, Qun Liu, Xiaoyan Zhu, and Min- lie Huang. 2023. DecompEval: Evaluating Generated Texts as Unsupervised Decomposed Question Answering. In ACL (1). Association for Computational Linguistics, 9676–9691

  4. [12]

    Keith and Amanda Stent

    Katherine A. Keith and Amanda Stent. 2019. Modeling Financial Analysts’ De- cision Making via the Pragmatics and Semantics of Earnings Calls. In ACL (1). Association for Computational Linguistics, 493–503

  5. [13]

    Subhendu Khatuya, Koushiki Sinha, Niloy Ganguly, Saptarshi Ghosh, and Pawan Goyal. 2024. Instruction-Guided Bullet Point Summarization of Long Financial Earnings Call Transcripts. In SIGIR. ACM, 2477–2481

  6. [14]

    Tushar Khot, Harsh Trivedi, Matthew Finlayson, Yao Fu, Kyle Richardson, Pe- ter Clark, and Ashish Sabharwal. 2023. Decomposed Prompting: A Modular Approach for Solving Complex Tasks. In ICLR. OpenReview.net

  7. [15]

    Mahoney, Kurt Keutzer, and Amir Gholami

    Sehoon Kim, Suhong Moon, Ryan Tabrizi, Nicholas Lee, Michael W. Mahoney, Kurt Keutzer, and Amir Gholami. 2024. An LLM Compiler for Parallel Function Calling. In ICML. OpenReview.net

  8. [16]

    JP Kincaid. 1975. Derivation of new readability formulas (automated readability index, fog count and flesch reading ease formula) for navy enlisted personnel. Chief of Naval Technical Training (1975)

  9. [17]

    Alice Kwak, Clayton Morrison, Derek Bambauer, and Mihai Surdeanu. 2024. Classify First, and Then Extract: Prompt Chaining Technique for Information Extraction. In Proceedings of the Natural Legal Language Processing Workshop 2024, Nikolaos Aletras, Ilias Chalkidis, Leslie Barr...

  10. [18]

    Patrick S. H. Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rock- täschel, Sebastian Riedel, and Douwe Kiela. 2020. Retrieval-Augmented Genera- tion for Knowledge-Intensive NLP Tasks. In NeurIPS

  11. [19]

    Guohao Li, Hasan Hammoud, Hani Itani, Dmitrii Khizbullin, and Bernard Ghanem. 2023. CAMEL: Communicative Agents for "Mind" Exploration of Large Language Model Society. In NeurIPS

  12. [20]

    Chin-Yew Lin. 2004. ROUGE: A Package for Automatic Evaluation of Summaries. In Text Summarization Branches Out. Association for Computational Linguistics, Barcelona, Spain, 74–81. https://aclanthology.org/W04-1013/

  13. [21]

    Mengpu Liu, Mengying Zhu, Xiuyuan Wang, Guofang Ma, Jianwei Yin, and Xiaolin Zheng. 2024. ECHO-GL: Earnings Calls-Driven Heterogeneous Graph Learning for Stock Movement Prediction. In AAAI. AAAI Press, 13972–13980

  14. [22]

    Yang Liu, Dan Iter, Yichong Xu, Shuohang Wang, Ruochen Xu, and Chenguang Zhu. 2023. G-Eval: NLG Evaluation using Gpt-4 with Better Human Alignment. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing. 2511–2522

  15. [23]

    Tim Loughran and Bill McDonald. 2014. Measuring readability in financial disclosures. the Journal of Finance 69, 4 (2014), 1643–1671

  16. [24]

    Sourav Medya, Mohammad Rasoolinejad, Yang Yang, and Brian Uzzi. 2022. An Exploratory Study of Stock Price Movements from Earnings Calls. In WWW (Companion Volume). ACM, 20–31

  17. [25]

    Rajdeep Mukherjee, Abhinav Bohra, Akash Banerjee, Soumya Sharma, Manjunath Hegde, Afreen Shaikh, Shivani Shrivastava, Koustuv Dasgupta, Niloy Ganguly, Saptarshi Ghosh, and Pawan Goyal. 2022. ECTSum: A New Benchmark Dataset For Bullet Point Summarization of Long Earnings Call T...

  18. [26]

    Timothy Nugent, George Gkotsis, and Jochen L. Leidner. 2023. Extractive Sum- marization of Financial Earnings Call Transcripts - Or: When GREP Beat BERT. In ECIR (2) (Lecture Notes in Computer Science, Vol. 13981) . Springer, 3–15

  19. [27]

    OpenAI. 2023. GPT-4 Technical Report. CoRR abs/2303.08774 (2023)

  20. [28]

    Yunxin Sang and Yang Bao. 2022. DialogueGAT: A Graph Attention Network for Financial Risk Prediction by Modeling the Dialogues in Earnings Conference Calls. In EMNLP (Findings). Association for Computational Linguistics, 1623–1633

  21. [29]

    R Senter. 1967. Automated readability index. (1967)

  22. [30]

    Karthik Suresh, Neeltje Kackar, Luke Schleck, and Cristiano Fanelli. 2024. To- wards a RAG-based Summarization Agent for the Electron-Ion Collider. CoRR abs/2403.15729 (2024)

  23. [31]

    Lei Wang, Chen Ma, Xueyang Feng, Zeyu Zhang, Hao Yang, Jingsen Zhang, Zhiyuan Chen, Jiakai Tang, Xu Chen, Yankai Lin, Wayne Xin Zhao, Zhewei Wei, and Jirong Wen. 2024. A survey on large language model based autonomous agents. Frontiers Comput. Sci. 18, 6 (2024), 186345

  24. [32]

    Qingyun Wu, Gagan Bansal, Jieyu Zhang, Yiran Wu, Beibin Li, Erkang Zhu, Li Jiang, Xiaoyun Zhang, Shaokun Zhang, Jiale Liu, Ahmed Hassan Awadallah, Ryen W White, Doug Burger, and Chi Wang. 2024. AutoGen: Enabling Next- Gen LLM Applications via Multi-Agent Conversations. In Firs...

  25. [33]

    Qingyun Wu, Gagan Bansal, Jieyu Zhang, Yiran Wu, Shaokun Zhang, Erkang Zhu, Beibin Li, Li Jiang, Xiaoyun Zhang, and Chi Wang. 2023. AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation Framework. CoRR abs/2308.08155 (2023)

  26. [34]

    Yiran Wu, Tianwei Yue, Shaokun Zhang, Chi Wang, and Qingyun Wu. 2024. State- Flow: Enhancing LLM Task-Solving through State-Driven Workflows. InFirst Con- ference on Language Modeling . https://openreview.net/forum?id=3nTbuygoop

  27. [35]

    Zhiheng Xi, Wenxiang Chen, Xin Guo, Wei He, Yiwen Ding, Boyang Hong, Ming Zhang, Junzhe Wang, Senjie Jin, Enyu Zhou, Rui Zheng, Xiaoran Fan, Xiao Wang, Limao Xiong, Yuhao Zhou, Weiran Wang, Changhao Jiang, Yicheng Zou, Xiangyang Liu, Zhangyue Yin, Shihan Dou, Rongxiang Weng, W...

  28. [36]

    Weinberger, and Yoav Artzi

    Tianyi Zhang, Varsha Kishore, Felix Wu, Kilian Q. Weinberger, and Yoav Artzi

  29. [2020]

    BERTScore: Evaluating Text Generation with BERT. In ICLR. OpenRe- view.net

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.